Main Dashboard
AI-Driven Insights
Weather Analysis
Order Tracking
Inventory
Kopik-AI Analysis

Kopi (Coffee in Indonesian) + K(nowledge)

Inspiration

Kopik was inspired by a close friend who runs a local cafe and constantly struggled with manual inventory decisions. Every morning meant guessing how much to order — often ignoring how weather spikes hot drink sales, or how nearby events suddenly double dessert demand. The result was wasted food on quiet days and frustrating stockouts during peak hours.

When we learned that the food industry wastes $162B annually while 76% of restaurants struggle with profitability due to poor inventory management, we knew this was a problem worth solving with technology.

What it does

Kopik is an AI-driven inventory management platform that helps cafes and bakeries make smarter ordering decisions by factoring in weather, local events, and historical sales.

Instead of guesswork, Kopik delivers actionable recommendations such as:

“Order 2 extra coffee bags for tomorrow’s rainy weather”
“Prepare 20% more desserts for tonight’s concert.”

The system integrates directly with POS platforms, runs AI analysis in real-time, and surfaces clear, prioritized insights through an intuitive dashboard — helping business owners reduce waste, prevent stockouts, and improve margins.

How we built it

Kopik was developed as a full-stack web application with a React frontend and FastAPI backend, supported by real-time data flows and AI agents.

Frontend

-Base 44 for the Frontend Development

React 19.1.1 + Vite for development speed and performance
React Router DOM for client-side navigation
Tailwind CSS for modern, responsive UI
Recharts + Framer Motion + Lucide React for interactive charts, animations, and icons
Centralized **React Context (DataContext) to manage live state from backend APIs

Backend

FastAPI with SQLAlchemy ORM for RESTful APIs
SQLite (kopik.db) for persistent storage
Pydantic v2 models for serialization and validation
AI System Uses GEMINI for real-time demand prediction and recommendation generation
REST endpoints under /api for inventory, recommendations, and analytics
Swagger Docs auto-generated for API exploration

Challenges we ran into

API architecture & data flow synchronization: The frontend was calling /intelligence/dashboard immediately after /intelligence/analyze without waiting for analysis to complete. This caused race conditions where stale data appeared while AI was still processing new inventory changes.
Real-time data integration: The system was mixing static JSON mock data with live API calls, leading to data inconsistencies, unpredictable behavior, and blocking scalable deployment.
Multi-system integration complexity: Connecting the React frontend, FastAPI backend, and AI agents required seamless communication between three different systems under tight hackathon deadlines while maintaining demo reliability.

Accomplishments that we're proud of

Built a scalable full-stack system with React + FastAPI in under 12 hours
Created a functional AI agent that analyzes inventory against external signals in real-time
Integrated live data flows: inventory changes immediately trigger AI analysis and dashboard updates
Designed a clean, modern dashboard that surfaces insights instead of overwhelming data

What we learned

Real-world inventory management is highly dynamic — demand forecasting must account for external context, not just past sales.
Small business owners value clarity and simplicity above raw AI power — actionable recommendations win over complex dashboards.
Building synchronous AI workflows requires balancing accuracy, latency, and system reliability.
We deepened our understanding of full-stack integration, combining React, FastAPI, SQLAlchemy, and Pydantic into a cohesive system.